Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Good Cooperation Between SAS and R for Regulatory Submission
Jingyuan Chen Hoffmann-La [email protected]
Objective
• Why to made SAS and R cooperated for submission work
• Illustration:
➢ Case one: to prove R can make submission ready products
o Example from eSUB Model 5
o Comparison
➢ Case two: SAS and R can complement each other for Exploratory analysis
o Exploratory analysis request from EMA
o Analysis Plan
o Demo involving R markdown
• Conclusion
Why to made SAS and R cooperated for submission work
• Open source becomes noticeable.
• SAS dominates the work of submission.
• Use the advantages from both, and use the right tool in right situation.
➢How much contribution it can make to this industry?
➢ Is R able to make submission ready products ?
➢How much effort it would take to do standard submission?
Illustration:Case one: use R for eSUB Model 5
• Background:
➢ To prepare e-submission to FDA second post marketing committee for a phase III oncology study with 516 patients.
➢ Module 5 : Clinical Study Reports▪ Listing of all clinical studies▪ Case report forms ▪ Study Reports▪ Datasets▪ Periodic Safety Update Reports▪ Literature References
Illustration:Case one: use R for eSUB Model 5
• Executing plan
➢ SAS contributed the majority part of Clinical Study Reports (CSR) and Analysis Datasets.
➢ Pilot R to generate two types of graphs, one table and one dataset.
➢ Working Process for R:
Illustration:Case one: use R for eSUB Model 5
• Quality Control Process
➢ Double program in traditional way, and applied company standard quality control process .
➢ Compare outputs from R and SAS.
➢ Use SAS program to compare submission dataset made by R and SAS.
Illustration:Case one: use R for eSUB Model 5
• Outputs comparison between SAS and R:
➢ Kaplan Meier Plot
Created using R
Created using SAS
CVP+Trt
CVP+Trt
Illustration:Case one: use R for eSUB Model 5
• Outputs comparison between SAS and R:
➢ Forest Plot
Created using R
Created using SAS
TreatmentTreatment
Illustration:Case one: use R for eSUB Model 5
• Outputs comparison between SAS and R:
➢ Survival Table
Created using R
Created using SAS
Treatment
Treatment
Illustration:Case one: use R for eSUB Model 5
• Outputs comparison between SAS and R:
➢ Laboratory Analysis Dataset
▪ Output data set is in .xpt format (SAS version 5)
▪ Used ‘Proc compare’ in SAS to compare the data produced by R and SAS
Illustration:Case one: use R for eSUB Model 5
• Program Efficiency Comparison between SAS and R
R SAS
Condition Calculation Loop
Take long time
If .. then in data step
Easy and fast
Change Data Structure
(wide format to long
format)
‘reshape’ function
Easy and fast
Transpose
Take some time
Select ‘last’ record ‘order’ and ‘aggregate’ function
Two lines of code
Sort and last in steps
Two data steps
Date format ‘as.Date’ function Input function
Format and Label ‘SASformat’ and ‘label’ function Format and label in data step
Xpt file ‘write.xport’ function
One line code, but take long
Libname xport and proc copy
Two steps
Illustration:Case two: cooperate SAS with R for Exploratory analysis
• Background:
➢ Trend observed towards an increased risk of early death for subjects treated with anti-PD-1/PD-L1
➢ European Medicines Agency (EMA) requested to conduct exploratory analyses aiming to identify factors that could predict the likelihood of not benefiting from immunotherapy.
➢ Complex issue and no established predictive factors have been identified so far.
▪ Include different products, indications, lines of treatment, biomarkers and diagnostic assays
▪ Different cut-offs used for study inclusion or as stratification factors.
Illustration:Case two: cooperate SAS with R for Exploratory analysis
• Analysis Plan
univariate model
Illustration:Case two: cooperate SAS with R for Exploratory analysis
• Analysis Execution
R Markdown
R
SAS
Illustration:Case two: cooperate SAS with R for Exploratory analysis
• Demo
15
Conclusion:
• R is capable to make submission standard products.
• Making the advantages from open source and SAS can speed up submission activities.
• Learning the features from different tools and using the right tool in the right situation
• Connect to new concepts, advanced theories, mature skills, and provide smarter analysis.
Acknowledgement
• Laura Harris
• Will Harris
• Aditi Qamra
Doing now what patients need next